The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes shell commands using find, awk, and bash to search for specific import patterns in the source code and tests. These are standard development operations for auditing a codebase.
[COMMAND_EXECUTION]: The test-side-effects.md file contains a Python snippet that uses subprocess.run to execute a small script in a fresh interpreter. This is a common and legitimate testing pattern used to verify that importing the library does not prematurely load heavy dependencies like PyTorch.
[EXTERNAL_DOWNLOADS]: The skill mentions common machine learning libraries such as torch, tensorflow, jax, and transformers as refactor targets. It also references uv and pytest for running tests. These are well-known, industry-standard tools and do not represent a security risk.
[PROMPT_INJECTION]: The instructions focus purely on code refactoring logic and do not contain any patterns typical of prompt injection or attempts to bypass agent safety filters.

lazy-import-refactor